Temperature calibration fluids

[DTJ]

Wasn't sure whether to post here or in Metrology.....

What would be a suitable fluid to submerge some bare temperature sensing PCBs into to force them all to the same ambient temperature?
I want them to be within 0.05 to 0.1°C of each other. The idea is to place the circuits in the fluid which is inside a well insulated enclosure and then wait until its reasonably stable.


The fluid would need to be:

1) non conductive.
2) easily removed (volatile?) - oils are no good as they are too messy and it will find its way under ic's etc.
3) easily procured.

Any ideas?

Isopropyl may not be good as it will be cooling as it evaporates perhaps causing issues of its own.

[sleemanj]

Is cost a factor?  If not, then maybe fluorinert.

[DTJ]

Is cost a factor?  If not, then maybe fluorinert.

Good idea, I'd forgotten all about Fluorinert.
Cost is a factor - Fluorinert is pretty costly, I found one place with 250mL  for US$199. If the boss was paying..... unfortunately I  am.

It looks like one of the Fluorinert family would do the job:
http://www.solvents.net.au/3m.htm

Hopefully someone will have some other ideas.

[babysitter]

There are cheaper candidates - silicone oil (e.g. for model car transmissions), simple petroleum, galden ls230 (you can use it for vapour phase soldering later) are candidates you might get smaller amounts much cheaper or use for something different.

Just make sure that it would be able to keep several sensors at equal temperature - cross-check with a single one and put it in different places.

[Marco]

There's some volatile silicone oils used in DIY cosmetics. Hydrophobic, volatile, easily available and cheap.

[ejeffrey]

How much heat do the boards generate?  If it is small enough you may not even need a liquid.  0.1 or even 0.05 C is achievable in air.  Put the boards in a thick walled ice cooler with a small fan (like a computer case fan) to keep the air well mixed.

Liquids are better of course.  But it is easy to try and doesn't require clean up.  If you needed 0.01 C or your circuits generate a lot of heat, I would say you need liquid cooling for sure.

If you can attach heat sinks to your PCB or sensors, that will help dramatically, particularly to reduce sensor self heating.  Of course if you can attach multiple sensors to the same aluminum bar or plate, that would be ideal.

Another thing that might make a difference is to add some thermal mass inside the cooler.  Put an inch of sand or gravel at the bottom before adding the boards and the fan.  That will take longer to equilibrate, but keep the temperature more stable over time.

[DTJ]

There are cheaper candidates - silicone oil (e.g. for model car transmissions), simple petroleum, galden ls230 (you can use it for vapour phase soldering later) are candidates you might get smaller amounts much cheaper or use for something different.

The Galden is interesting. I had a quick look to see where I could buy it from in Au, it still seems pretty expensive. I'll look at it more, someone like Sparkfun might be selling it.

There's some volatile silicone oils used in DIY cosmetics. Hydrophobic, volatile, easily available and cheap.

Thanks Marco - this looks very interesting. I found some "Cyclomethicone Volatile Silicone Oil Cosmetic Grade" on eBay for $40 / litre. It looks very promising. My only concern that may be unfounded is that it may leave some residue behind when evaporating which would stop the potting compound from bonding properly later on (unless I wash the pcbs yet again). It's readily available too in smaller volumes.
https://www.ebay.com.au/itm/Cyclomethicone-Volatile-Silicone-Oil-Cosmetic-Grade-Shine-Silk-Conditioning/142312233830?hash=item2122785b66:m:meELrIZfv67gBDb4qWROWWg

How much heat do the boards generate?  If it is small enough you may not even need a liquid.  0.1 or even 0.05 C is achievable in air.  Put the boards in a thick walled ice cooler with a small fan (like a computer case fan) to keep the air well mixed.

Liquids are better of course.  But it is easy to try and doesn't require clean up.  If you needed 0.01 C or your circuits generate a lot of heat, I would say you need liquid cooling for sure.

If you can attach heat sinks to your PCB or sensors, that will help dramatically, particularly to reduce sensor self heating.  Of course if you can attach multiple sensors to the same aluminum bar or plate, that would be ideal.

Another thing that might make a difference is to add some thermal mass inside the cooler.  Put an inch of sand or gravel at the bottom before adding the boards and the fan.  That will take longer to equilibrate, but keep the temperature more stable over time.

Luckily the boards generate very little heat (effectively zero).

I thought that placing them in an insulated container would work if the sensors were very close together (ie within 3mm or so). I did not try mixing the air. With 5 sensors stacked with a 3mm gap I get a variation of up to 0.5°C between them (no particular pattern).

However if I place two PCBs together with a small brass cylinder thermally bonding the (QFN16) sensors together  and wrap the assembly in bubble wrap then I get readings that match to <0.05°C within 20 seconds or so.

Thanks for the ideas.

[jbb]

I hear from experienced mechies and production engineers that silicone residues can be very annoying and stop adhesives from bonding properly.

Maybe a thick copper/aluminium plate for thermal mass and a nice squishy thermal gap pad to bring the components to temperature?

[beanflying]

I have the bits together to make a wet calibration well for RTD's, Thermocouples and other sensors.

Air is a terrible medium for heat transfer you also need to design carefully for keeping the chamber homogeneous but then with stirring or airflow comes weird variations or without comes temperature gradients (neither are great) so I would suggest you are never going to get near the sorts of accuracy you are chasing using air.

Will take some photos tomorrow of the bits I have but for a chamber I have imported a 2L twin walled s/steel benti (food) box off aliexpress. The wells are some very phallic objects

The basic plan is 4 wire RTD as a reference with up to 6 other wells around it. Crude version in absolute terms search for "fluke triple point cell" to get a reference at 0.00 deg C. These wells would contain IPA for most readings but as they are capped and submerged in an ice bath air given time and suitable insulation on top would still give some good readings or the well could even be packed with a non electrically conductive substance (silica or sand?). Due to the nature of the vessel adding an element to the bottom and some control should allow multi point calibration using deminerilised R/O water to hit a 100 degree point and checking along the way.

In theory it could be driven below zero and well above depending on the medium chosen in the vessel. I have been looking at some silicone oils for the higher temp purpose.

[beanflying]

Couple of photos of the bits. Plan is to use one or possibly both of the inner trays to tig the tubes in place with a layer of foam between. Lid is screw top but some sort of gland or outlet/inlet for cables will be needed with another layer of insulation as the tops of the tubes will be here.

And the phallic bits are alcohol flasks. Would you drink out of one of these in public? 

[DTJ]

I hear from experienced mechies and production engineers that silicone residues can be very annoying and stop adhesives from bonding properly.

Maybe a thick copper/aluminium plate for thermal mass and a nice squishy thermal gap pad to bring the components to temperature?

"Yep silicone can cause all sorts of grief for painting so I imagine it would be the same for potting compounds.

I've got some copper bus bar in the workshop, I'll give that a go. I think the thermal pad might be useful too as I have noticed when there is a difference between a couple of sensors that moving it around and re-seating usually eliminates the error.

[DTJ]

I have the bits together to make a wet calibration well for RTD's, Thermocouples and other sensors.

Air is a terrible medium for heat transfer you also need to design carefully for keeping the chamber homogeneous but then with stirring or airflow comes weird variations or without comes temperature gradients (neither are great) so I would suggest you are never going to get near the sorts of accuracy you are chasing using air.

Will take some photos tomorrow of the bits I have but for a chamber I have imported a 2L twin walled s/steel benti (food) box off aliexpress. The wells are some very phallic objects

The basic plan is 4 wire RTD as a reference with up to 6 other wells around it. Crude version in absolute terms search for "fluke triple point cell" to get a reference at 0.00 deg C. These wells would contain IPA for most readings but as they are capped and submerged in an ice bath air given time and suitable insulation on top would still give some good readings or the well could even be packed with a non electrically conductive substance (silica or sand?). Due to the nature of the vessel adding an element to the bottom and some control should allow multi point calibration using deminerilised R/O water to hit a 100 degree point and checking along the way.

Thanks for going to the effort to post the pics.

I looked at the dual walled bento boxes on ebay last night - quite interesting. I've been using a stainless thermos flask but the neck is a little too narrow.

I watched a video on youtube for an icepoint set up and they had a 'self draining' dewar - a dual walled set up with a drain hole through the bottom to let out excess melt water.

I also found online the manual for a fluke calibration bath that was a good read.
https://www.google.com.au/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&uact=8&ved=0ahUKEwj8vL70l8zXAhUEGpQKHcLiAdsQFggmMAA&url=http%3A%2F%2Fdownload.flukecal.com%2Fpub%2Fliterature%2F7911A__2ugeng0000.pdf&usg=AOvVaw0FtGXQvcis5w2vfHJm90s9


What size batteries do those flasks take? C or D? 

[beanflying]

What size batteries do those flasks take? C or D? 

Mmmmm Buzzz - Pass I don't know 

[Marco]

I hear from experienced mechies and production engineers that silicone residues can be very annoying and stop adhesives from bonding properly.

How many will have had an application for a volatile silicone compound though?

[jbb]

How many will have had an application for a volatile silicone compound though?

Good point. Basically none of them, I suspect.

Still, I would try for a dry process if possible; easier to control and less waste/cleanup.

[DTJ]

The outcome!

I ended up using clothes pegs to clamp the PCBs to a 1/4" x 2" x 5" piece of copper bus bar.
The QFN packaged sensor IC is pushed against the copper. I lay some bubble wrap over the whole thing and leave it sit for 5 minutes to stablilise.


With a row of 10 PCBs spread across about 12cm they all read within about  +/- 0.05°C

Adjacent units read within about 0.02°C. If I see an odd one, re-clamping it seems to bring it closer to the average.

I'll have to find some heat-sink silicone sheet and see if that helps.


Thanks for all the ideas - they'll be stored away for use another day.